Summary: | 碩士 === 國立成功大學 === 測量及空間資訊學系 === 106 === Building footprint is an essential part of urban environmental studies. Object-Based Image Analysis (OBIA) can be utilized to efficiently extract the building footprint from the Digital Surface Model (DSM) and true-orthoimage especially for the large urban area. The procedure of OBIA is to segment the image into image objects and classify them into different land covers according to object features. Then, the building footprint can be extracted from the building class. However, the building footprint is jagged and irregular because of image segmentation. Furthermore, the original straight edges and right-angles of building footprint may disappear. Thus, in this research, we propose an algorithm to regularize the building footprint. The study area is located in Taipei City. There are various types of buildings which can be used to evaluate the potential of the proposed method. The building density in our study area is very high and many buildings are connected to each other. It is a difficult challenge, even the commercial software cannot handle it, such as ArcGIS. Hence, the shared edges of adjacent buildings should be handled separately to keep the topology correct and must not have overlap and gap between buildings. The Douglas-Peucker algorithm is utilized to simplify the building footprint. The complexity of footprint features will be reduced and be more concise. After simplification, the building footprint needs to be regularized. Most of the buildings are rectangular which have the same direction to the road nearby. Therefore, the principal orientations of buildings can be determined based on road vector to assist the building footprint regularization. Moreover, the shared edges are simplified and regularized only once and applied for the polygons on both sides, which can avoid the topology errors. Through the proposed approach, the regularized building footprint will be the straight line with right-angle which are regular and reasonable.
|